Google has evolved Stitch into an AI-native software design canvas that lets anyone create, iterate and collaborate to turn natural language into high-fidelity UI designs. The redesigned tool, announced this week from Google Labs, adds voice interaction and a design philosophy that challenges how teams approach building interfaces.
The new canvas features an AI-native, infinite workspace that gives ideas room to grow from early ideations to working prototypes. Rather than spending hours sketching wireframes, users can now describe what they want their audience to feel and let the AI generate multiple design directions that match that vibe.
The headline feature is voice control. Voice Canvas lets you speak directly to your canvas, with the AI agent listening, asking clarifying questions, giving real-time design critiques, and making live updates. Users can request things like three different menu layouts or ask to see a screen in various colour palettes, watching the changes happen as they speak. This removes the friction of clicking through menus or typing precise prompts when exploring early design directions.
Vibe design replaces the traditional wireframe-first workflow, allowing you to describe a business objective, a desired user feeling, or a design inspiration, with Stitch generating multiple design directions that match that vibe. This represents a shift from pixel-perfect specificity to intent-based creation. A founder might say 'make this feel premium and minimal, like Stripe' or 'this needs to feel urgent but not pushy.' Stitch takes that direction and builds variations.
Once designs are polished on the canvas, the tool connects to downstream tools in a development workflow. Using the Stitch MCP server and SDK, users can leverage Stitch's capabilities via other tools or export designs to developer tools like AI Studio and Antigravity, ensuring partnership between designers, AI, and developers remains seamless. The biggest leap is the ability to generate a fully functional React application from selected screens, not just a clickable prototype, but actual working code exportable to AI Studio or other tools.
The sceptic's question is obvious: does this actually work at speed? Google suggests Stitch helps professional designers exploring dozens of variations or founders visualising their first software idea get work done in minutes rather than days. That claim sits in a long tradition of design tool marketing; the real test is whether the output holds up under production scrutiny.
The tool's greatest strength is speed of exploration; you can go from a vague idea to a clickable multi-screen prototype faster than you could set up a new Figma file. For founders still validating ideas or teams exploring multiple design directions quickly, that matters. Its greatest weakness is depth; for production design systems, team collaboration, and pixel-perfect refinement, Figma and similar tools remain essential, with Stitch best understood as the beginning of the design process, not the end.
Stitch is live at stitch.withgoogle.com for users 18 and older in every region where Gemini is available. The tool is free, a significant advantage against paid competitors. Users get 350 generations per month in Standard mode powered by Gemini 2.5 Flash, and 50 in Experimental mode with Gemini 2.5 Pro.
The broader question Stitch raises is whether design and development are converging. For product teams, Stitch could compress timelines that currently span weeks into hours, with the bottleneck between 'we need this feature' and 'here's a clickable prototype' always being designer availability and iteration cycles; if Stitch delivers on its promises, it might let engineers and product managers mock up interfaces directly, saving designers for higher-level system thinking rather than pixel-pushing. That's the premise. Whether execution matches intent will take months of real-world use to determine.